Lightning Talk: Agentic Bee: How To Get AI Agents To Talk To Tetragon
summary
Connects Tetragon runtime telemetry to LLM agents via a simple RAG architecture and FlowPulse sensors over Tetragon’s gRPC API to push dynamic tracing and enforcement policies. Uses Sift + Grype + CVE lookups to map files to vulnerabilities, then GPT-4 and Claude Sonnet 3.0 agents build behavioral evidence chains (JSON, images, text) to prioritize and act. A demo shows it reduces 700 vuln to 3 and generates container reports in 1–2 minutes, 30k–60k tokens, ~$0.25 per report.
AI-written summary · about
chapters (11)
- 0:00Agentic Bee and Tetragon intro
- 0:22High-fidelity observability for AI agents
- 1:26High-fidelity data layer with Tetragon
- 1:47Raw data to asset evidences
- 2:29GPT-4 and Claude agent workflows
- 2:50Vulnerability prioritization workflow
- 4:59Threat detection multi-crew workflow
- 7:08Vulnerability prioritization demo
- 8:13Threat hunting and IR demo
- 9:36Exporting evidences to platforms
- 10:18Q&A
related talks
01h 07mEmpowering Security Teams with Generative AI
24mAI Agent Swarms Are Amazing
49mAI Agents: Collaboration in Real-Life Applications
33mWhat Did My Agent Do? Observability and Accountability for AI Agents
39mTowards Useful, Reliable, and Scalable AI Agents
19mAI Agents for Incidents: The Good, the Bad, and the Ugly
30mBuilding Reliable AI Agents: An Open Source Approach To Evaluation and Observability
10mLightning Talk: Building AI That Ops Teams Actually Trust